Show simple document record

Other titre : Transform- and statistical-based image analysis for assessment of deterioration in concrete infrastructure

dc.contributor.advisorRivard, Patricefr
dc.contributor.advisorHe, Dong-Chenfr
dc.contributor.authorKabir, Shahidfr
dc.date.accessioned2014-05-15T12:32:42Z
dc.date.available2014-05-15T12:32:42Z
dc.date.created2008fr
dc.date.issued2008fr
dc.identifier.urihttp://savoirs.usherbrooke.ca/handle/11143/1819
dc.description.abstractThe evaluation of the condition of infrastructure requires the development and optimization of alternative inspection methods for assessing surface deterioration in order to obtain accurate and quantitative information to supplement visual inspections. To this end, non-destructive methods that produce image data show great potential and are increasingly being used in concrete applications. These methods, based on IR thermography and greyscale and colour imaging, are generally cost and time effective and are relatively easy to employ. There is, however, a need for efficient image analysis techniques to extract relevant damage information from the imagery. In this context, this research proposes the use of grey level co-occurrence matrix statistical texture analysis in combination with Haar's wavelet analysis for improved defect detection. Two classifiers are proposed, the supervised multi-layer perceptron artificial neural network and the unsupervised K -means clustering approach, for evaluation of their effectiveness in characterizing the deterioration from the imagery. These techniques are applied to thermographic, colour and greyscale imagery of laboratory specimens and field samples exhibiting different levels of concrete deterioration. Further experiments are conducted on borehole acoustic imagery involving the additional techniques of spatial filters and edge-detectors in an effort to determine their efficiency in detecting concrete damage.The results demonstrate that the hybrid texture approach is quite effective for defect discrimination. They also indicate that the lowpass and median spatial filters performed better than the gradient-based and Laplacian edge-detectors; however, the texture approaches outperformed all of the other techniques.The artificial neural network was found to provide better classification accuracies compared with the K -means algorithm. Concerning the imagery, the thermography produced more accurate results than the colour and greyscale imagery.The information derived from the imagery consists of total surface damage; for map-crack imagery, the total length of cracks and range of crack width openings were also computed.The damage quantities obtained for the laboratory specimens show good correlation with test measurements recorded for the specimens, such as expansion and impact-echo velocities.fr
dc.language.isoengfr
dc.publisherUniversité de Sherbrookefr
dc.rights© Shahid Kabirfr
dc.titleMéthodes d'analyse d'images pour l'évaluation de la dégradation des structures en bétonfr
dc.title.alternativeTransform- and statistical-based image analysis for assessment of deterioration in concrete infrastructurefr
dc.typeThèsefr
tme.degree.disciplineGénie civilfr
tme.degree.grantorFaculté de géniefr
tme.degree.levelDoctoratfr
tme.degree.namePh.D.fr


Files in this document

Thumbnail

This document appears in the following Collection(s)

Show simple document record